Introduction
Ecosystem Services (ES), the benefits people obtain from ecosystems, are fundamental for human well-being. Scientific interest has increased substantially over the last decades, particularly since the Millennium Ecosystem Assessment (MA) in 2005. ES (or Nature’s Contributions to People (NCP) in IPBES) have become central concepts for linking biodiversity, ecosystem condition and human well-being and for providing information for sustainability policy and decision-making, from local to international scales (Díaz et al. 2015, Pascual et al. 2017, Díaz et al. 2018).
Over the past decade, several conceptual frameworks and assessment guidelines, such as the System of Environmental-Economic Accounting – Ecosystem Accounting (SEEA-EA) (United Nations et al. 2024), the European Union-initiated Mapping and Assessment of Ecosystems and their Services (MAES) framework (Maes et al. 2020) and the Intergovernmental Science-Policy Platform on Biodiversity and Ecosystem Services (IPBES) (Díaz et al. 2015), have consolidated the use of ES/NCP in ecosystem assessments. These frameworks highlighted the need for assessment approaches that are conceptually robust, comparable across different realms, and useful for decision support (Balvanera et al. 2022). They place increasing demands on the description, mapping, monitoring and valuation of direct and indirect contributions of ecosystems to human activities, linking, among others, ecosystem type (ET), ecosystem condition and ES supply in a coherent analytical chain (Jones et al. 2016, Kadykalo et al. 2019, Nicholson Thomas et al. 2025). However, operational guidance for determining when and how natural and anthropogenic contributions should be made explicit in ES assessments remains limited.
A long-standing assumption in ES science has treated ES as arising solely from natural capital, with humans positioned largely as passive beneficiaries. This framing has been progressively replaced by an understanding of ES/NCP as products of social–ecological interactions, in which natural capital and a range of anthropogenic inputs jointly shape what is ultimately experienced or used by people (Fischer and Eastwood 2016, Palomo et al. 2016, Díaz et al. 2018). The IPBES conceptual framework explicitly acknowledges that anthropogenic assets – together with institutions and governance – mediate the realisation and distribution of NCP (Díaz et al. 2015, Pascual et al. 2017). Many generated ES flows are the result of ecological structures, processes and functions combined with human-derived assets, capabilities and practices, such as built infrastructure, technology, knowledge, institutions, and financial resources. For instance, crop production depends on soils, fertilisation, seeds, pollinators, climate and water alongside labour, machinery, agronomic knowledge and market access. Flood regulation through nature-based solutions such as terraces or restored wetlands combines ecological functions with engineering, maintenance and governance arrangements (Palomo et al. 2016, Lavorel et al. 2020, Mansoldo et al. 2026); recreation and other cultural ES are mediated by access infrastructure, interpretation, identity and social practice (Raymond et al. 2018). Even in apparently "natural" systems (e.g., forests), the realised ES flows frequently result from human contributions that determine whether ecosystem potential is mobilised, transformed or sustained over time (Jones et al. 2016).
Co-production has emerged as the conceptual lens through which these joint dependencies are analysed, but it is far from being a unified concept. A recent critical review identified several overlapping framings of ES/NCP co-production – input, actor, and context-focused, and broader disciplinary perspectives rooted in sustainability science, political ecology, post-humanist scholarship and institutional analysis, or relational values – each highlighting different facets of the human-nature interaction (Woodhead et al. 2025). Input-focused framings emphasise the types and quantities of natural and anthropogenic capital mobilised in producing a service, whereas actor- and context-focused framings stress histories and places, who participates, through which institutions, and under which values and power relations (Fischer and Eastwood 2016, Rodela et al. 2019, Isaac et al. 2022, Locatelli et al. 2024, Woodhead et al. 2025). Thus, co-production is used differently across research domains. Each framing highlights distinct dimensions of social–ecological systems, and should therefore be understood as complementary rather than competing.
This work adopts an input-focused and assessment-oriented framing, drawing on the capital-based perspective (Palomo et al. 2016, Jones et al. 2016, Lavorel et al. 2020). Co-production is therefore conceptualised as the joint contribution of natural and anthropogenic inputs to the generation of realised ES/NCP flows. Natural capital is understood here as the interacting biotic and abiotic components of ecosystems, including ecological structures, processes, functions and condition, that underpin ES/NCP. Anthropogenic inputs serve as an umbrella term for human, social, physical and financial capital. Human capital includes knowledge, skills, labour and capabilities; social capital includes networks, trust, institutions, norms and collective capacities; physical capital includes infrastructure, equipment and technology; and financial capital includes monetary resources that enable investment, maintenance and use (Goodwin 2003). Operationally, co-production includes anthropogenic inputs that shape ecosystem potential or capacity, as well as inputs that mobilise, access, use, experience or transform realised ES/NCP flows. This narrowing is analytical rather than normative: the aim is not to reduce ES/NCP to inputs alone, but to render natural and anthropogenic contributions explicit and, where feasible, measurable, thereby improving assessment quality and its relevance for resource-use decisions.
Despite broad conceptual recognition, empirical operationalisation of co-production remains uneven, inconsistent and difficult to assess. Studies differ in terminology, indicators, methods, spatial and temporal scales, ET and ES categories considered (Kachler et al. 2023, Woodhead et al. 2025). Evidence of co-production is concentrated on croplands, managed forests and material or provisioning ES, while regulating and cultural ES, and many marine, freshwater and urban contexts, remain comparatively under-investigated (Kachler et al. 2023). Physical and human capital, being more visible and measurable, tend to dominate empirical accounts, whereas social, institutional and financial contributions are often acknowledged but their quantification is overlooked (Isaac et al. 2022, Kachler et al. 2023). Distinguishing natural and anthropogenic contributions is conceptually challenging in managed and hybrid systems, where nature-based solutions are applied, and in cultural ES where meaning and access are partly constitutive of the service itself (Raymond et al. 2018, Schutter et al. 2021, Locatelli et al. 2024). Locatelli et al. (2024) distinguish ecosystem-level mechanisms, through which anthropogenic inputs shape ecosystem potential, from flow-level mechanisms, through which they mobilise or transform realised contributions, and underline that natural and anthropogenic inputs may be complementary, substitutable or interdependent. Full causal decomposition is therefore rarely feasible and not always desirable; what is required, instead, is an appropriate basis for deciding when explicit treatment of ES co-production is necessary. The methodological gap addressed here is therefore not the absence of methods for measuring co-production per se, but the lack of a transparent decision logic for determining when such assessment is necessary and for selecting a fit-for-purpose methodological approach.
This requirement is particularly relevant for integrated ecosystem assessments, which seek to move beyond approaches based only on ET and ecosystem extent by linking drivers of change, pressures and ecosystem condition with ES potential, realised flows, beneficiaries, values and decisions within an analytical chain (Jones et al. 2016, Balvanera et al. 2022, Seguin et al. 2026). More recently, the Framework for Integrated Ecosystem Assessment (FIEA) further formalised this chain-based view, emphasising that ecosystem condition, ES flows, values and decision contexts should be treated as connected components of the assessment process (Lange et al. 2026). The distinction between ecosystem potential or capacity and realised flow is therefore fundamental. Ecosystem potential refers to the ability of an ecosystem, given its condition, to generate ES before they are mobilised, accessed, transformed or used by people. Thus, ES potential describes the hypothetical maximum yield of a particular ES (Burkhard et al. 2012). Realised flow or use refers to the actual contribution mobilised, accessed, experienced, transformed or consumed by people after relevant anthropogenic inputs have been applied. Jones et al. (2016) argue that potential ES, realised ES and value capture different aspects of ES supply and should not be conflated. This is critical because the same realised ES flow can be produced through different and complex combinations of ecosystem condition and anthropogenic inputs, with heterogenous implications for sustainability assessment, trade-off analysis, ecosystem accounting, restoration planning, nature-based solutions design, environmental justice and decision-making. High crop yields may reflect intensive use of fertilisers and pesticides, irrigation and mechanisation rather than good ecosystem condition. Flood regulation may result from ecological functions combined with engineered infrastructure and governance arrangements. Cultural ES supply (e.g., nature-based recreation) may depend on ecological features, but also on access, interpretation, identity, memory and social inclusion (Palomo et al. 2016, Lavorel et al. 2020). Failure to make these distinctions can misattribute outcomes to ecosystems and obscure the human inputs, on which realised flows depend, weakening both ecological and economic accountability and the equity of benefit distribution (Pascual et al. 2017, Isaac et al. 2022, Ojeda et al. 2022).
The present study responds to this gap by developing a decision-oriented basis for determining when and where explicit consideration of ES co-production is warranted in integrated ecosystem assessments, and for guiding the selection of relevant anthropogenic inputs and appropriate assessment approaches across ecosystem types and services. The contribution lies in linking existing approaches with ET × ES relevance screening and fit-for-purpose method selection in a transparent decision sequence. Specifically, it combines: (i) a structured synthesis of the conceptual and methodological literature on ES/NCP co-production; (ii) an expert-elicited relevance matrix across ET × ES combinations; and (iii) a set of evidence-based case-study examples to illustrate operationalisation in different ecosystem types and decision-making contexts. The study focuses on realised ES flows while preserving the analytical distinction between ecosystem potential or capacity and realised flows, which underpins ecosystem accounting and integrated ecosystem assessment (Jones et al. 2016, Balvanera et al. 2022).
The analysis is guided by three research questions:
For which ET × ES combinations is an explicit consideration of anthropogenic inputs most relevant when assessing realised ES flows?
Which forms of natural and anthropogenic capital are most important across different ET × ES contexts, and where do they enter the ES assessment chain?
Which assessment approaches and indicators are most suitable for operationalising co-production across different ecosystem types, services, spatiotemporal scales and decision-making contexts?
Literature review: an evidence synthesis
Rationale for a structured narrative synthesis
This Section synthesises the conceptual and methodological literature on ES/NCP co-production. It is organised as a structured narrative synthesis rather than a systematic review. The aim is not to consider all studies on ES/NCP co-production, nor to suggest the prevalence of methods or framings, but to assemble the body of conceptual, empirical and methodological work that can guide upcoming co-production studies in the framework of integrated ecosystem assessments. Three recent reviews provide the evidentiary backbone for this approach. Kachler et al. (2023) mapped empirical evidence on ES/NCP co-production and showed that studies have unevenly represented ET and anthropogenic capital types. Woodhead et al. (2025) reviewed how ES co-production is framed in literature and identified several distinct research traditions. Locatelli et al. (2024) refined the theoretical basis of co-production by distinguishing ecosystem-level and flow-level mechanisms and by clarifying the roles of complementarity and substitutability between natural and anthropogenic inputs. These review papers provide a foundation, but they do not provide an integrated decision logic for screening when explicit co-production assessment is warranted and selecting a fit-for-purpose methodological approach. The intention here is not to duplicate these reviews, but to organise their findings, together with previous conceptual frameworks (Palomo et al. 2016, Fischer and Eastwood 2016, Jones et al. 2016, Lavorel et al. 2020) and selected methodological applications. Thus, the aim is to facilitate the identification of methods that are suitable for different assessment purposes, rather than to rank them, considering heterogeneous ecosystem, service and decision-making contexts.
Conceptual frameworks
The literature on ES/NCP co-production has developed through several, partially overlapping, framings. Although there is convergence on the recognition that ES/NCP arise from social–ecological interactions, literature diverges in how the contribution of anthropogenic capital, including knowledge and assets, is framed. The IPBES conceptual framework explicitly recognises that human well-being depends not only on nature, but also on anthropogenic assets, institutions, governance systems and diverse knowledge systems (Díaz et al. 2015). Further NCP framings emphasised the role of culture, knowledge plurality and multiple value systems in mediating people–nature relationships (Pascual et al. 2017, Díaz et al. 2018). Although ES and NCP differ in conceptual lineage and terminology, both concepts are highly relevant for integrated ecosystem assessments and both require evaluating how ecological structures and processes contribute to human well-being under specific social, economic and institutional conditions (Kadykalo et al. 2019). Accordingly, we use ES/NCP when referring generically to both concepts, while retaining ES or NCP where a specific framework, classification, analytical construct or cited source uses one term explicitly. The main framings are summarised in Table 1 and are considered complementary rather than mutually exclusive; they show different aspects of the same social–ecological reality and their selection consequently determines the choice of an appropriate analytical method.
Capital, cascade/chain-based and ecosystem/flow-level co-production
Within an evolving group of framings, co-production is conceptualised as the joint mobilisation of natural capital and one or more forms of anthropogenic capital. This provides a starting point for integrated assessment because it follows an analytical chain logic and provides a vocabulary for distinguishing ecological from human-derived inputs (e.g. which capitals are required, where they enter the chain and whether they are measurable). Palomo et al. (2016) describe co-production as a continuum and identify pathways through which different capitals shape realised ES flows, with implications for sustainability (e.g. quantity and quality of services), trade-offs, resilience and equity. Jones et al. (2016) develop a closely-related framing and extend this logic in terms of stocks and flows of natural and human-derived capital, distinguishing potential service from realised service and explicitly placing human inputs at the points where ecosystem outputs are mobilised, transformed or accessed; ES supply can occur at multiple stages, i.e. shaping potential service supply, mediating demand and enabling the realisation of flows.
A closely-related framing is the cascade- or chain-based approach, where co-production is not treated as a single event, but as a sequence through which ecosystem condition and potential are transformed into realised flows and benefits (Lavorel et al. 2020); the sequence follows the logic of ecosystem condition, potential/capacity, management/shaping, mobilisation/access, realised flow, benefit/value and distribution and feedbacks. When it comes to integrated assessments, this approach can be quite useful, as it clarifies that anthropogenic inputs may act before, during or after the generation of ecosystem potential. For example, restoration and land management may shape flood regulation; infrastructure and institutions may mobilise that potential; and social access, risk perception or governance may determine whether the realised benefit is equitably distributed. Thus, anthropogenic inputs shape ecosystem potential through management, enable flows through infrastructure or labour or mediate benefit through access, institutions and values.
Locatelli et al. (2024) articulate this further by distinguishing co-production at the ecosystem level from the flow level. In the former, human actions influence the ecological structures and processes that generate ecosystem potential or capacity, for example, through land-use planning, restoration, conservation management, fertilisation, irrigation or mulching. In the latter, human actions mobilise, access or transform that potential into realised ES/NCP flows and associated benefits, for example, through harvesting, infrastructure use, access provision, interpretation, labour or institutions. This distinction is central for operationalisation because it clarifies whether anthropogenic inputs act by modifying ecosystem potential, enabling realised flows or mediating access to benefits. The relationship between natural and anthropogenic inputs is, therefore, context-dependent: in some cases, anthropogenic inputs may partially compensate for or substitute natural inputs, particularly where ecosystems are degraded or intensively managed; in others, natural and anthropogenic inputs are complementary, interdependent or essential and non-substitutable. These different relationships have important implications for whether ecosystem and human contributions can be analytically separated, compared or jointly represented.
Human–nature interaction and meaning-based framings
Human–nature interaction and meaning-based framings focus on the relational, embodied and practice-based dimensions of co-production. Fischer and Eastwood (2016) propose an analytical framework in which co-production emerges from interactions between people's identities, capabilities and practices and the ecological systems they engage with; in short, ES arise through interactions between people and place, while the framework distinguishes amongst the co-production of ecosystem structures, the ES co-production and the attribution of meaning. This understanding is consistent with broader accounts of culture as skilled, situated engagement with particular environments rather than as a fixed set of traits or values (Ingold 2021). Raymond et al. (2018) similarly emphasise that, for cultural ES, meaning, interpretation and embodied experience are partly constitutive of service realisation rather than merely post hoc perceptions attached to a prior biophysical flow. These framings are particularly important for tangible and intangible (non-material) cultural ES and for situations where human agency is not adequately represented as a separable input, i.e. where realised flows may depend on identities, capabilities, embodied practices and social interpretation rather than solely on biophysical supply. Additionally, such framings prevent an overly narrow treatment of co-production as a technical input bundle, highlighting why some anthropogenic contributions – such as knowledge, belonging, social practices, cultural interpretation and access capabilities – cannot be adequately represented by physical or monetary indicators alone.
Governance, access and value-mediated framings
These framings highlight that institutions, property regimes, access rules, power asymmetries and value pluralism mediate which contributions are realised and by whom. Pascual et al. (2017) argue that valuation methods must accommodate plural values to be policy-relevant. Ojeda et al. (2022) reframe people–nature relationships as reciprocal, drawing attention to the contributions people make to nature as well as those they receive. Isaac et al. (2022) developed a governance perspective on the management of multiple anthropogenic capitals in ES/NCP co-production. They showed that governance systems influence the management of anthropogenic capitals underpinning co-production across spatial, administrative and temporal scales. This is consistent with perspectives that conceptualise adaptive governance as a dynamic, learning-orientated and cross-scale process that shapes the conditions under which social–ecological systems are managed and transformed (Folke et al. 2005). They also identify a gap in how governance is incorporated into co-production research. This framing can be particularly important considering that realised ES/NCP flows depend not only on ecological potential and material inputs, but also on property rights, collective rules, knowledge systems, power relationships, access arrangements and distribution. Dean et al. (2021), Barraclough et al. (2022), Mayer et al. (2025)and related studies point to the importance of stakeholder networks, institutions, property rights, social relationships and cross-scale dependencies. In integrated ecosystem assessments, these factors can be particularly relevant, for example, where the same ecological potential produces different realised benefits for different groups or, to put it differently, where realised flows depend less on simple biophysical supply and more on those who can access, manage, mobilise or benefit from ecosystem contributions.
Accounting-boundary framings
This framing is associated with SEEA-EA (United Nations et al. 2024), which defines ES as the contributions of ecosystems to benefits used in economic and other human activity. In the accounting process, there is a “recognition” that many benefits are produced jointly by ecosystems and human inputs such as labour, produced assets, intermediate inputs, technology, infrastructure and leisure time. Although the accounting task is not meant to represent co-production as a full multi-capital process, it isolates the ecosystem’s contribution to avoid double counting and maintain consistency with national accounts. Thus, although it is not a co-production framework per se, ecosystem accounting can be relevant for co-production, as it formalises the boundary between ecosystem contribution and benefit. However, it limits the degree to which anthropogenic capitals are analysed jointly with natural capital within the ES account itself, even though produced assets are represented elsewhere within the wider System of National Accounts. Woodhead et al. (2025) draw out the contrast between this boundary-based logic and more inclusive input-, actor- and context-focused framings, emphasising that operational choices about scope and boundaries are themselves theoretically loaded.
Method families
Operationalising ES/NCP co-production requires methods that can identify, distinguish, model, map or account for the respective roles of natural capital and ecosystem properties, human inputs, management practices, institutions and beneficiaries in generating realised ES/NCP flows. As co-production occurs across the ES/NCP chain – from ecosystem condition and potential supply, through management, mobilisation and access, to realised flows, benefits and values – no single method can capture all relevant mechanisms. The most appropriate method depends on the assessment function, ecosystem context, data availability, indicator suitability, scale and decision purpose (Nicholson Thomas et al. 2025). Based on selected empirical studies, the methodological literature is organised in this Section into six families, distinguished broadly by what they attempt to disentangle, i.e. capital inputs, spatial flows, marginal contributions, management/scenario mechanisms, stakeholder meanings and roles and governance/network relations.
Indicator-, capital- and typology-based approaches operationalise co-production through proxies for natural capital, anthropogenic inputs, management intensity, landscape structure and ES outputs (Balvanera et al. 2022). These approaches are useful for screening and structured comparison across sites, management types or landscape units within a defined assessment context. In agroecosystems, Bethwell et al. (2021) showed that provisioning ES cannot be adequately indicated by agricultural yield alone, because outputs depend on site conditions, anthropogenic co-generation and external environmental effects. Similarly, Torralba et al. (2018) used surveys, indicators and multivariate analysis to classify Spanish dehesa systems and relate management models to ES bundles, while Wolff et al. (2021) used plot-level agricultural data and cluster analysis to identify agricultural landscape types differentiated by structure, diversity and management intensity. The strengths of these approaches lie in transparency and scalability, while their main limitation is that indicators often remain static proxies (Palomo et al. 2016, Jones et al. 2016), for example, they may show that natural and anthropogenic capitals co-occur, but not whether one causes, substitutes for or complements the other.
Spatial ES-flow, serviceshed and supply–use mapping approaches disentangle co-production by identifying where ES are generated and who benefits. They are particularly relevant for ES whose realised flows depend on spatial relationships amongst service-providing areas, service-benefitting areas and connecting mechanisms, including pollination, recreation, flood and erosion regulation, water and food supply. In an accounting context, Remme et al. (2018) provide a spatially explicit national-scale biophysical supply–use account for the Netherlands, with results also reported at the provincial level, modelling 13 ES and allocating physical ES flows to ecosystem types and economic users. La Notte (2022) adds an agri-food example by showing that crop provision can be assessed either as an aggregated ecological input or through a more disaggregated treatment of supporting ecological processes, such as erosion regulation and crop pollination. Treating intermediate ecological processes as additional final ES can lead to double-counting, whereas a disaggregated approach can clarify the specific contribution and trends of individual ES flows within agricultural production systems. These approaches are well-suited to ecosystem accounting, although institutions, power relationships and informal access mechanisms are often represented only indirectly. Cascade-based approaches can extend this logic by making explicit the transition from potential supply to value realisation. Broome et al. (2024), for example, expand the ES cascade for auroral ES by distinguishing co-production processes, such as value attribution, mobilisation of ES potential, appropriation and commercialisation, thereby showing how demand-side action and socio-technical systems are required for some benefits to materialise.
Production-function, econometric and capital-substitution approaches estimate realised ES flows or benefits as functions of ecosystem inputs and anthropogenic inputs. Their main advantage is that they can quantify marginal contributions, elasticities or relative input effects where outputs and inputs are measurable. Grammatikopoulou et al. (2020) provide an example by estimating the marginal productivity and implicit value of green water in Czech cereal production using a Cobb–Douglas production-function approach. These approaches are strongest for provisioning ES and some measurable regulating ES, but they are data-intensive, sensitive to model specification and less suitable for cultural, relational and access-mediated ES/NCP. Broader modelling literature also shows why co-production should not be interpreted simply as substitutability. Rieb et al. (2023) use a simulation model to show that the long-term substitutability of natural and human-derived capital depends on how capitals change over time and how they interact; reproducing a near-term function of natural capital with human-derived capital does not necessarily mean that substitution is viable in the long term.
Process-based, management, scenario and trajectory approaches represent the ecological, management and temporal mechanisms through which co-production changes under alternative conditions. Crop, hydrological, soil, carbon, erosion, flood and habitat models can represent ecosystem condition, potential supply and management effects, while scenario and counterfactual designs can compare degraded versus restored systems, high- versus low-input management or conventional versus conservation agriculture. Historical-trajectory approaches add a temporal dimension by examining how co-production changes as social–ecological systems adapt to policy, market, climate and demographic drivers. Bruley et al. (2021) provided an example from the Pays de la Meije in the French Alps, where literature, archives, interviews and participatory workshops were combined in a chronosystemic timeline to identify four historical periods of change and to analyse how ES/NCP, adaptive responses and natural, human, social, financial and manufactured capitals were reconfigured over time. Evidence synthesis could also be grouped in this methodological family. Tamburini et al. (2020), for instance, showed through a second-order meta-analysis of diversification practices that agricultural diversification generally enhances multiple ES, including pollination, pest control, nutrient cycling, soil fertility and water regulation, often without compromising crop yields.
Participatory, qualitative and stakeholder-elicitation approaches examine co-production as a social–ecological process involving knowledge, practices, values, perceived responsibilities, legitimacy and access. These approaches are essential where co-production depends on meanings, identities, local ecological knowledge, place attachment, stakeholder roles or perceived quality of life. Dean et al. (2021) translated ES evidence on pastoral systems and showed that habitat creation and maintenance, food and feed and supporting identities are strongly connected to social–ecological systems; material, regulating and non-material contributions are closely linked to mobility, local knowledge, identity and long-term human management. Giacomelli et al. (2024) used focus groups and interviews in Le Marche, Italy, to examine stakeholder roles in ES co-production and to distinguish the contributions of human, social, physical and financial capital across ES categories. Klaus et al. (2025) showed that participatory approaches are not only diagnostic, but also necessary for implementation: balancing grassland ES requires multi-stakeholder processes to assess societal ES demand, design multifunctional landscapes, motivate farmers and implement ES-enhancing actions across field, farm and landscape scales. These approaches provide depth and contextual validity; however, they are less standardised and less directly transferable across regions than indicator-based or statistical methods.
Governance, network, value-realisation and benefit-sharing approaches focus on the actors, institutions, rules, collaborations, power relations and distributional mechanisms that determine whether ecosystem potential is mobilised into realised benefits. They are particularly important where ES delivery depends on collective action, cross-scale governance, property rights, access arrangements, social–ecological fit, social trust or contested values. Barraclough et al. (2022) provided a mixed-methods example from the Nordhordland UNESCO Biosphere Reserve in Norway, combining PPGIS, focus groups, surveys and social network analysis to map stakeholder–ES relationships, governance networks and mismatches between valued ES and institutional priorities. Zhu et al. (2025) extend this governance perspective towards value realisation by developing an ecological value creation framework for the Xinghua Duotion ecosystem in China. Their framework links ES co-production to ecological value management, ecological value cognition and ecological benefit-sharing and shows that equitable allocation of rights and responsibilities, local ecological-cultural awareness, social trust and benefit-sharing mechanisms enable the realisation of ES value. Systems methods such as fuzzy cognitive mapping can further represent feedbacks, perceived causal links and trade-offs, as shown by Heindorf et al. (2025) in peatland co-production scenarios. These methods are highly relevant for decision support and governance diagnosis, although their outputs are usually less directly compatible with ecosystem accounting.
Methodology
The previous Sections have provided the conceptual and methodological basis for the empirical design. As the relevance of co-production varies across ET × ES combinations and because dominant capitals and assessment-relevant chain stages vary across contexts, this study combines two complementary components: (i) an expert-elicited ET × ES relevance matrix; and (ii) a set of case-study examples illustrating how co-production can be operationalised across different ecosystem types and methodological approaches. The matrix is used as a screening device to identify where explicit treatment of anthropogenic inputs is likely to add value; it is not intended to measure the magnitude of co-production or causally apportion realised ES flows between natural and anthropogenic inputs. Where more explicit treatment is warranted, the literature synthesis and case-study examples provide information for consideration of the relevant anthropogenic capitals and the stages at which they enter the ES assessment chain and provide guidance on suitable methodological pathways. These two components are described below.
Expert-elicited co-production relevance matrix
To elicit expert judgement on the relevance of anthropogenic inputs when assessing realised ES flows, a structured ET × ES relevance matrix was developed within a dedicated working group on ES co-production, established under the SELINA EU Horizon project (https://project-selina.eu/). The columns represented 12 ETs following the MAES classification (Maes et al. 2020): urban; cropland; grassland; woodland and forest; heathland and shrub; sparsely vegetated land; wetlands; rivers and lakes/inland surface waters; marine inlets and transitional waters; coastal systems; shelf systems; and open ocean. The rows initially comprised 26 ES, categorised into provisioning, regulating and maintenance, cultural and appreciation services, following the SEEA-EA reference ES list (United Nations et al. 2024). The three residual “other” summary ES categories were excluded from active scoring, as their broad and insufficiently specified scope would have reduced the robustness and interpretability of expert judgement; they were retained only to preserve taxonomy traceability. The remaining 23 ES formed the actively scored set. Each scorable ET × ES combination contained two response fields: a relevance score and a corresponding confidence score. In addition, each expert declared a single level of expertise for each ET.
Following the distinction between ecosystem-level and flow-level co-production (Locatelli et al. 2024), the elicitation focused on the relevance of anthropogenic inputs for realised ES flows, rather than for ES potential or capacity alone. Before scoring, the matrix was pre-screened by the working-group leadership team to identify ET × ES combinations that were not suitable for realised-flow assessment under the adopted framing. Combinations were structurally excluded where no plausible realised flow could be specified in the reference context or where the available conceptual or empirical basis was insufficient to support a meaningful expert judgement. Such combinations include soil-related regulating services in open-ocean or shelf systems and flood-mitigation combinations where no plausible ES flow pathway could be specified within the adopted ET boundary. The 12 ETs combined with the 23 ES generated 276 possible ET × ES combinations, of which 45 were excluded during the pre-screening. The remaining 231 combinations formed the analytical universe for expert assessment and subsequent analysis.
The matrix was completed independently by 16 experts from the previously described SELINA ES co-production working group. Experts were recruited from the project consortium through a purposive and voluntary-participation process. As the purpose was to elicit structured expert judgement, rather than to infer results to a wider population, the panel is best understood as an informed rather than statistically representative sample of the wider ES research community. To support transparency, the survey recorded expert metadata, including affiliation, disciplinary background, role, years of experience and primary ET expertise (Suppl. material 2). Prior to scoring, the working group met regularly over a six-month period to discuss definitions, applications and methodological issues related to co-production. This process established a shared conceptual foundation, following matrix-based expert elicitation practice (Burkhard et al. 2009, Campagne and Roche 2018), while maintaining independent judgement at the scoring stage.
Experts were asked to score all non-excluded ET × ES combinations, beginning with the ETs corresponding to their primary expertise and then proceeding to the remaining cells. Judgements could be informed by direct case-study experience or by broader conceptual knowledge. To maintain a common reference frame, all scores referred to an ecosystem context within Europe. For each combination, experts were asked: “Based on your knowledge, how relevant are anthropogenic inputs for realised ES flows in this ET × ES combination in a European context”? Responses were recorded on a six-point relevance scale: 0 = not relevant, 1 = very low relevance, 2 = low relevance, 3 = medium relevance, 4 = high relevance and 5 = very high relevance. A high score, therefore, indicated that anthropogenic inputs make an important contribution to the realised ES flow and should be explicitly considered in respective integrated ecosystem assessments.
For each response, experts also recorded their confidence in the relevance score on a five-level scale, from 1 = very low confidence to 5 = very high confidence. This score captured the certainty of the expert’s own judgement for that specific ET × ES combination. Experts further declared their level of expertise for each ET on a four-level scale: 0 = no specific expertise, 1 = limited familiarity, 2 = good familiarity and 3 = strong expertise. This expertise score was recorded at the ET level and applied to all ES within that ET.
After the first round of independent scoring, experts were shown the aggregated panel results and were invited to review their judgements. Experts could revise their scores where they considered this appropriate. This step followed the consensus-round logic of the Delphi method (Jacobs et al. 2015) and aimed to support convergence of judgements without imposing consensus. The results presented below use the final scores, while individual expert consistency was examined as a diagnostic check (Table S10, Suppl. material 1).
The aggregation of scores followed a two-stage procedure. The first stage estimated the panel’s relevance verdict for each assessed ET × ES combination, while the second assessed the degree of trust that could be placed in that verdict. This separation ensured that relevance and trust were not conflated. For each assessed combination, the mean relevance score across experts was calculated and subsequently converted into a relevance verdict described below (Table S1, Suppl. material 1). The Standard Deviation (SD) of the relevance scores was used as a measure of agreement amongst experts (Table S3, Suppl. material 1). The mean confidence score was also calculated for each combination (Table S4, Suppl. material 1), while the mean expertise score was calculated for each ET (Table S5, Suppl. material 1). Consistent with cautions in the matrix-method literature, relevance scores were not weighted by self-reported confidence, as this can amplify overconfidence bias (Campagne et al. 2017, Campagne and Roche 2018). Confidence and expertise were, therefore, used only to qualify the trust placed in the relevance verdict, as described below. Median relevance was retained as a complementary robustness metric, but was not used as the verdict.
The mean relevance score was converted into a three-level relevance category using fixed cut-points. ET × ES combinations with a mean relevance score > 3.5 were classified as “Relevant”, those with a mean score < 1.5 were classified as “Not relevant” and those between these thresholds were classified as “Intermediate”. These cut-points are classification conventions used to translate aggregated expert relevance scores into screening categories rather than empirically or ecologically validated thresholds, i.e. they distinguish relevance judgements for screening and prioritisation purposes. Trust in each relevance verdict was assessed independently using two dimensions: the strength of the evidence supporting the judgement and the level of agreement amongst experts. Evidence (E) was calculated as the geometric mean of normalised mean expertise and normalised mean confidence, as follows:
\documentclass[12pt]{standalone}
\usepackage{varwidth}
\usepackage[utf8x]{inputenc}
\usepackage[T1]{fontenc}
\usepackage{lmodern}
\usepackage{amsmath, amssymb, graphics, setspace}
\newcommand{\mathsym}[1]{{}}
\newcommand{\unicode}[1]{{}}
\newcounter{mathematicapage}
\begin{document}
\begin{varwidth}{50in}
\begin{equation*}
E = \sqrt{\frac{\bar{e}}{3} \times \frac{\bar{c}-1}{4}}
\end{equation*}
\end{varwidth}
\end{document}
(1)
where ē is the mean expertise score and c̄ is the mean confidence score. E ranges from 0 to 1. The geometric mean ensures that evidence is high only when both expertise and confidence are high; if either component is low, the evidence score is reduced. Agreement was represented by the SD of the relevance scores. Evidence and agreement were then each grouped into three levels. Evidence was classified as Limited (E < 0.40), Medium (0.40 ≤ E ≤ 0.70) or Robust (E > 0.70). Agreement was classified as High (SD ≤ 1.0), Medium (1.0 < SD < 1.5) or Low (SD > 1.5). These two classifications were crossed in a 3 × 3 lookup table, adapted from Jacobs et al. (2015), to assign one of five trust tiers, ranging from Very low to Very high (Fig. 1).
Before assigning final relevance and trust categories, a knowledge-gap rule was applied to determine whether the panel had sufficient expertise for each ET. This rule took precedence over the remaining procedure: if expertise was insufficient for an ET, all ET × ES combinations for that ET would be labelled as “Knowledge gap” and no relevance verdict would be assigned. In the reported analysis, the rule was applied leniently. An ET was retained if at least one respondent declared the highest expertise level, i.e. 3 = strong expertise. This minimum threshold reflected the matrix’s purpose as an exploratory screening and prioritisation tool rather than a basis for statistical inference. A more restrictive exclusion rule would have removed potentially informative judgements for under-represented ETs; limited expert coverage was instead reflected in the mean-expertise component of the trust assessment and sensitivity analysis (Table S9, Suppl. material 1). Accordingly, relevance verdicts for ETs with limited strong-expertise coverage should be interpreted with caution.
Each assessed ET × ES combination received a composite score combining the relevance category and the trust tier, for example, “Relevant – Very high” and was represented by a corresponding traffic-light colour in the matrix. As a diagnostic check, a confidence- and expertise-weighted mean relevance score was calculated alongside the standard unweighted mean and the difference between them was reported for each combination (Table S2, Suppl. material 1). This diagnostic identified ET × ES combinations where better-informed experts diverged from the panel as a whole, but it was not included in the final relevance verdict to avoid double-counting confidence and expertise in both the relevance score and the trust assessment. All thresholds, including the relevance cut-points, evidence and agreement bands and knowledge-gap rule, were varied in a sensitivity analysis to assess the stability of the resulting classifications.
Selection of co-production case-study examples
To complement the matrix assessment, five case-study examples were analysed to illustrate how co-production can be operationalised in empirical assessments. The purpose was not to validate or reject the matrix results, nor to provide a statistically representative sample of European ecosystem assessments. Rather, the cases were used as evidence-rich examples of how different methods, indicators or decision-support approaches can make natural and anthropogenic contributions to realised ES flows visible in specific social–ecological and policy contexts.
The case studies were identified through the SELINA co-production working group. Contributors were members of the working group with direct involvement in and detailed knowledge of the selected cases. Cases were considered suitable where sufficient documentation of the co-production assessment logic was available. Selection also sought diversity across ET, ES and assessment purposes, so that the cases could illustrate different routes for implementing co-production rather than repeat a single methodological design.
A structured template was used to screen and analyse suitable cases within the working group. The template followed the capital- and chain-based framing (Palomo et al. 2016, Jones et al. 2016, Lavorel et al. 2020) and was implemented through a two-part questionnaire (Suppl. material 3). The first part addressed the main forms of capital involved in co-production. Contributors were asked to identify the relevant inputs and, where possible, indicate whether these acted directly or indirectly in relation to the focal ES. The second part addressed assessment methods and decision-making context, including models, tools, indicators, data sources, spatial and temporal scale, treatment of anthropogenic inputs and relevance for planning, management or policy decisions. Contributors could provide separate responses for multiple ES within the same case, where this was appropriate.
The collected information was compiled and cross-checked to retain a common conceptual structure while drawing on the strongest available evidence. Contributors of the selected case studies were then asked to complete a short case-study summary template and provide any available documentation for each case, such as peer-reviewed publications, reports, project deliverables or planning documents (Suppl. material 4). The five selected cases were analysed according to how each assessment represented: (i) ecosystem potential or capacity; (ii) anthropogenic inputs; (iii) realised ES flows; (iv) environmental externalities, benefits, values or trade-offs where available; and (v) the extent to which natural and anthropogenic contributions were separated, compared, jointly represented or qualitatively interpreted.
Table 2 summarises the selected case studies, including their ET × ES coverage, main assessment approaches and methodological value for analysing co-production, while Fig. 2 shows their geographic locations and provides a visual overview of the contrasting social–ecological contexts represented.
Results
Expert-elicited co-production relevance across ecosystem types and services
Fig. 3 presents the final co-production relevance matrix for the 231 assessed ET × ES combinations, based on the aggregated judgements of 16 experts. The matrix combines relevance and trust in a traffic-light scheme: colour hue indicates the relevance verdict (red = not relevant, orange/yellow = intermediate, green = relevant), while colour intensity indicates the trust tier. Structurally excluded combinations are shown in dark grey, while flat grey was reserved for knowledge gaps. Under the adopted knowledge-gap rule, all 12 ETs were retained and no assessed combination was classified as a knowledge gap.
Across the matrix, Intermediate was the most frequent verdict, accounting for 125 of the 231 combinations, compared with 65 Relevant and 41 Not relevant combinations. This pattern indicates that, within this panel and reference context, anthropogenic inputs were judged to have at least some relevance for most assessed ET × ES combinations, although often not decisively so. Trust was also concentrated in the middle of the scale, with Medium as the modal tier (115 combinations), followed by 77 combinations at High or Very high trust and 39 combinations at Low or Very low trust. The complete distribution and per-combination statistics are provided in Table S8 (Suppl. material 1).
Relevance varied across ETs (Table S6, Suppl. material 1). Anthropogenic inputs were judged consistently relevant in ETs characterised by intensive land use or direct management. Urban had the highest mean relevance score (3.92), with 18 of the 23 ES classified as Relevant, followed by Cropland (3.68), with 17 of the 23 ES classified as Relevant. Together, Urban and Cropland accounted for 35 of the 65 Relevant ET × ES combinations. Most other ETs were dominated by Intermediate verdicts, including Grassland (2.93), Woodland and Forest (2.88), Rivers and Lakes (2.63), Wetlands (2.57), Marine Inlets and Transitional Waters (2.51) and Coastal systems (2.48). The lowest relevance scores were observed for less-intensively managed ETs, i.e. Sparsely Vegetated Land (1.93), Shelf systems (1.75) and Open Ocean (1.71). Together, these three ETs accounted for 22 of the 41 Not relevant combinations.
Relevance also differed across ES (Table S7, Suppl. material 1). Cultural and appreciation services recorded the highest scores. Education, scientific and research services had the highest mean relevance score (3.95) and were classified as Relevant in every ET in which they were assessed. Other high-scoring services included Recreation-related services (3.69), Ecosystem and species appreciation (3.62) and Spiritual, artistic and symbolic services (3.32). Provisioning services, together with soil-, water- and biotic-interaction-related regulating and maintenance services, occupied the upper-middle range, including Flood mitigation (3.08), Water flow regulation (3.07), Biomass provisioning (2.94), Pollination (2.90), Soil and sediment retention (2.88) and Water supply (2.84). By contrast, the lowest scores were found for atmospheric and broad-scale climate-regulating services, particularly Rainfall pattern regulation (1.38), Air filtration (1.55), Local climate regulation (1.76) and Global climate regulation (1.82), for which anthropogenic inputs were generally judged to have limited relevance for realised ES flows.
The strongest verdicts occurred where high-relevance ES coincided with high-relevance ETs. The 14 combinations classified as Relevant – Very high were concentrated in Cropland and in selected cultural and recreational services of Urban and Woodland and Forest ETs. In Cropland, these included Biomass provisioning (mean relevance = 4.94), Soil quality regulation, Soil and sediment retention, Pollination, Biological control, Nursery population and habitat maintenance and Water flow regulation ES. In Urban ET class, relevant combinations included Recreation-related ES, Visual amenity, Education, scientific and research and Flood mitigation ES; in Woodland and Forest ET class, they were limited to Education, scientific and research ES and Ecosystem and species appreciation ES. By contrast, the strongest Not relevant verdicts were mainly found for atmospheric or climate-regulating ES in Sparsely vegetated, Coastal, Inland- or Open-water ETs, such as Air filtration and Local climate regulation ES.
Trust in the verdicts broadly reflected the panel’s coverage of each ET. Higher-trust verdicts were concentrated in terrestrial ETs, whereas marine and sparsely vegetated ETs generally received lower-trust verdicts, reflecting thinner declared expertise for these ETs. In particular, mean expertise for Shelf, Open Ocean, Marine Inlets and Coastal systems was in the range of 0.6–1.2. The relevance categories were nevertheless robust to plausible changes in the classification thresholds, with only a minority of combinations reclassified to an adjacent category when the cut-points were varied (Table S9, Suppl. material 1). The knowledge-gap outcome, however, was sensitive to the screening rule itself. Requiring at least two experts at the highest level rather than one would reclassify Shelf – the only ET retained on the basis of a single such expert – and all 13 of its assessed combinations as knowledge gaps, while leaving all other ETs unaffected. Accordingly, Shelf verdicts should be interpreted with caution and jointly with their trust tiers and sensitivity results; under the stricter expert-coverage rule, the appropriate classification for all assessed Shelf combinations would be a knowledge gap.
Case studies
The following case studies complement the expert-elicited matrix by illustrating how co-production can be operationalised in empirical assessments and how different methods can make anthropogenic inputs to realised ES flows visible in specific decision-making contexts. The five cases cover contrasting systems (Table 2) and collectively illustrate different analytical entry points for assessing co-production, depending on whether the main question concerns economic viability, management intensity, restoration prioritisation, urban risk reduction or the role of institutions and land-tenure regimes.
Troodos Mountains, Cyprus: comparative cost–benefit assessment in mountain terraced agroecosystems
The Cyprus case illustrates co-production through a comparative assessment of a nature-based solution (i.e. terracing) under contrasting agricultural management regimes. In the Troodos Mountains, viticulture has historically depended on traditional drystone terraces constructed with basic tools and family labour. These terraces transformed steep slopes into cultivable land and supported multiple ES, including biomass provisioning (e.g. through viticulture), soil and sediment retention, water flow regulation, visual amenity and spiritual, artistic and symbolic services, as well as associated mountain livelihoods. However, demographic change, land fragmentation, high labour requirements and declining maintenance have contributed to terrace abandonment, wall collapse, soil erosion and loss of productive land (Zoumides et al. 2017). The case compares the traditional system with modern, mechanically constructed terraces, created using earth-moving machinery and occasionally reinforced by newly-built drystone walls. Modern terraces allow wider benches, tractor access, drip irrigation and precision viticulture practices.
The co-production mechanism is assessed not only as a biophysical process, but also as an investment and justice question. Terracing modifies ecosystem potential at slope scale by reshaping runoff pathways, soil stability and cultivable land availability, while infrastructure, irrigation, labour and wine processing convert that potential into realised provisioning and cultural-economic benefits. In this respect, terraced agroecosystems are shaped by mountain biophysical conditions, ecological processes, human knowledge, built structures and governance (Zoumides et al. 2025).
The case combines Cost–Benefit Analysis (CBA) with a social justice enquiry and is, therefore, best placed within the process-based, management/scenario and decision-support family of methods, with additional relevance to governance and benefit-distribution approaches (see literature review section). The CBA compared modern, mechanically constructed terraces combined with agroecological vineyard management and wine production against continued use and minimal maintenance of traditional vineyard terraces over a 15-year horizon (Giannakis et al. 2025). Monetary flows were expressed per planted hectare, discounted at 6% and assessed using Net Present Value (NPV) and Benefit–Cost Ratio (BCR). The analysis incorporated production costs, terrace construction and maintenance costs, vineyard infrastructure, labour, subsidies, yield progression and market revenues. Specifically, the modern system generated an NPV of €48,456/ha and BCR of 1.10, compared with an NPV of – €4,807/ha and BCR of 0.82 for the traditional system under the baseline assumptions (Giannakis et al. 2025).
The assessment revealed several forms of anthropogenic contribution. Some inputs, such as machinery, terrace construction costs, farm labour, irrigation infrastructure and market valorisation, can be represented in monetary terms (Giannakis et al. 2025). Other co-produced ES, such as those related to cultural landscape values (e.g. visual amenity and spiritual, artistic and symbolic services) and regulating services (e.g. water flow regulation and soil and sediment retention), are more difficult to monetise fully and may remain only partly captured in conventional financial indicators (Zoumides et al. 2022, Giannakis et al. 2026). For instance, erosion reduction can be quantified biophysically through field monitoring, modelling or combined qualitative-quantitative methods (Djuma et al. 2017, Camera et al. 2018); translating these regulating ES effects into monetary values of co-produced benefits, however, remains methodologically demanding. Additionally, traditional and modern terrace systems embody different ecological, cultural and aesthetic values that may shape social acceptance, landscape identity and restoration legitimacy, but are only partially represented in conventional CBA. The case therefore shows both the usefulness and the limits of CBA for co-production assessment, i.e. such methods can clarify the economic viability of alternative management regimes, but conclusions remain sensitive to site-specific assumptions on yields, farming costs, subsidies, discount rates and market prices and do not fully capture the multidimensional benefits and trade-offs associated with the complexity of (traditional) mountain farming systems.
Northern German croplands: enhanced indicator framework for cropland provisioning services
The German case provides a clear input–output indicator example amongst the five cases. Its starting point is the insufficiency of agricultural yield as a proxy for provisioning ES supply in intensively managed croplands. Bethwell et al. (2021) applied an enhanced indicator framework in three northern German agricultural regions that differ in natural conditions and production structures: Diepholz has relatively high soil quality, higher precipitation, livestock production and biogas; Uelzen is characterised by irrigated potato and sugar beet production; and Oder-Spree has lower precipitation, poorer soils, limited irrigation and a focus on cereals and rapeseed. The focal realised ES flow was biomass provision from agricultural systems, while the wider framework also considered soil- and water-related environmental externalities associated with provisioning ES supply.
The case-study explicitly separated ecosystem potential, anthropogenic inputs, realised ES flows, environmental externalities and demand or preference dimensions. Natural capital was represented through site conditions such as soil quality and agroclimatic conditions, which define agroecosystem potential. Anthropogenic inputs included fertilisers, pesticides, fuel, irrigation water, machinery, labour and management decisions. These inputs act primarily at the flow level by mobilising and transforming ecosystem potential into crop and livestock outputs, but they may also feed back into ecosystem-level conditions through nutrient balances, greenhouse gas emissions, erosion risk and other externalities.
Methodologically, the case belongs primarily to the indicator-, capital- and typology-based family, complemented by bio-economic farm modelling. The bio-economic farm model MODAM was used to simulate agricultural decision-making and income optimisation under region-specific production structures. It incorporated production practices, agricultural inputs, labour and machinery data and was parameterised using statistical data and interviews with farmers and regional experts. The indicator set then grouped co-production-relevant information into several analytical classes. Ecosystem potential was represented by the M-SQR soil quality rating index (Mueller et al. 2007). Direct anthropogenic inputs were quantified through biophysical and monetary variables, including seeds, fertilisers, pesticides, fuel, irrigation water, working time and machine use. Aggregated and relational indicators included nutrient inputs, standardised treatment index, total fuel use, total factor costs and nitrogen balances. Realised ES flow was measured through crop yields, crop sales, grain-equivalent units and monetary values, while environmental externalities were represented through greenhouse gas emissions and selected input-related pressure indicators.
The results illustrate why co-production matters analytically in croplands. Soil quality ratings provided a useful indication of ecosystem potential; potential alone, however, did not explain realised flows or farm income. High outputs in Diepholz were associated with high nitrogen inputs and labour, whereas Uelzen showed higher pesticide treatment, fuel use and irrigation water due to its potato and sugar beet farming systems. Biophysical and monetary indicators also led to different interpretations: Diepholz showed high biophysical output, while Uelzen achieved higher income through higher gross margins and long-term investments in irrigation and marketing infrastructure. The case, therefore, shows that realised provisioning ES depends not only on ecological potential, but also on management intensity, production systems, market access, capital investment and historical development. The methodological challenges are equally informative: while such input-focused assessment can make the co-production of realised provisioning ES more transparent, it is time- and data-intensive, farm- and site-specific. Thus, disentangling natural and anthropogenic contributions is most appropriate where available data and decision needs justify the additional analytical effort.
Peloponnese, Greece: spatial prioritisation for post-fire forest restoration
The Greek case addresses a different decision-making problem: how to prioritise restoration or natural recovery in fire-affected Mediterranean landscapes. It focuses on post-fire restoration planning for several burnt natural areas affected by wildfires in 2021 and 2022. Amongst others, the assessment included wildfire-affected sites in the Peloponnese, i.e. Messinia–Arkadia, Ileia–Arkadia, Portes Achaia and Krestena (Tzamtzis et al. 2022, Tzamtzis et al. 2023). The focal ES are mainly regulating and maintenance services, including soil and sediment retention (including landslide/mass-movement mitigation), nursery population and habitat maintenance, water purification and global climate regulation, together with cultural ES as nature-based recreation services and visual amenity.
The contribution to co-production assessment lies in treating human intervention as a targeted and evidence-based input, rather than a generic restoration effort. Natural capital is represented by vegetation type, fire-adaptive traits, regeneration potential, slope intensity, unburnt patches and ecological features of affected Natura 2000 sites, including protected habitat types, species assemblages and conservation-relevant vegetation structures. Human and social capitals enter through post-fire ecological documentation and assessment, Forest Service responsibilities, protected-area governance, Natura 2000 designation and associated protection rules, legal designation of reforested areas, grazing restrictions and transparent prioritisation under conditions of limited financial and human resources and public pressure for extensive reforestation (Tzamtzis et al. 2023). Manufactured and financial capital are less directly measured, but are implicit in erosion-preventative measures, restoration plans, planting material, spatial data infrastructures and the allocation of restoration resources.
The method followed in this case is best understood as a spatial decision-support approach within the process-based, management/scenario and governance-orientated families. Tzamtzis et al. (2023) developed a systematic approach for recording, mapping, assessing and prioritising burnt natural areas in Greece. The method uses transparent criteria to determine whether active restoration or natural recovery is more appropriate, focusing on the fire-adaptive capacity of affected forest species, fire history or recurrence and slope steepness. Burnt ecosystems lacking fire-adaptive mechanisms, such as fir or black pine forests, are prioritised for active restoration, while fire-adapted pine systems (e.g. Pinus halepensis, Pinus brutia forests) are assessed according to whether repeated burning may have exhausted regeneration potential and whether steep slopes increase erosion risk. The relevant baseline report similarly combined burnt-area mapping, forest-map development, vegetation-unit identification, 20-year fire-overlap mapping, slope criteria and targeted field inspections to justify and prioritise human-induced reforestation, considering previous forest-management legacies that have shaped the pre-fire ecosystem condition and ES supply potential (Tzamtzis et al. 2022).
The case links co-production directly to implementation. Anthropogenic inputs include erosion-preventative measures before winter rainfall, grazing restrictions to protect post-fire regeneration, selection of native species and planting material and specialised management plans for logging or silvicultural interventions. These actions are intended to safeguard future ES flows by supporting ecosystem recovery and reducing post-fire degradation. The approach also connects restoration planning to biodiversity and conservation policy, as some of the studied fires affected Natura 2000 protected areas, making restoration prioritisation relevant to conservation targets and EU nature restoration objectives (Tzamtzis et al. 2023). Importantly, the spatial decision-tree can make co-production operational and transparent, but its transferability depends on the quality and compatibility of geospatial datasets, vegetation maps, fire-history records and local expert validation (these data are available nationwide for Greece). Similarly, the criteria are, in principle, transferable across Mediterranean fire-prone landscapes, but require adaptation where species traits, fire regimes and institutional responsibilities differ.
Trento, Italy: urban green infrastructure for heat and runoff regulation
The Italian case illustrates co-production in an urban social–ecological–technical system, where regulating ES depend on interactions amongst vegetation, built morphology, drainage infrastructure and technical management. Trento is a medium-sized city in northern Italy, located on the floodplain of the Adige River and surrounded by forested slopes, agricultural land, river corridors and semi-natural areas. Urban expansion, transport infrastructure, soil sealing and ecological fragmentation have placed pressure on the city’s green and blue infrastructure. In the valley-floor context, limited air circulation and dense human activity intensify summer heat stress, while high imperviousness increases pluvial flooding risks. The ES assessment was developed to support the city’s Urban Greening Plan, which frames green and blue infrastructure as core for biodiversity, climate adaptation, ecological connectivity and citizens’ well-being (Comune di Trento 2024). The focal ES are local climate regulation, water flow regulation and flood mitigation.
Co-production is operationalised through high-resolution spatial modelling of urban regulating services. Natural capital includes tree canopy, vegetation structure, permeable soils, riparian corridors, the Adige River and tributaries and the surrounding forested landscape. These components contribute to shading, evapotranspiration, infiltration, runoff attenuation and ecological connectivity. Anthropogenic inputs include buildings, roads, drainage systems, green-space design and maintenance. Planning rules, tree inventories, monitoring infrastructure and spatial decision-support tools are treated as enabling institutional, informational and technical capacities. Human and social capitals are represented by expertise in urban ecology, hydrology, hydraulic engineering, meteorology, landscape planning, urban forestry, geospatial analysis and municipal green-space management, as well as by the co-creation process underpinning the Urban Greening Plan (Comune di Trento 2024).
Geneletti et al. (2025) describe the use of multi-scale integrated modelling in Trento to support ecological planning strategies linked to climate-change resilience. Water flow regulation and flood mitigation (stormwater runoff) were assessed using the Storm Water Management Model (SWMM), combining hydrological modelling of runoff generation with hydraulic routing through the urban drainage network. Inputs included land-cover data, a digital terrain model, rainfall data for an extreme pluvial event, urban drainage elements and sub-catchments. Indicators included runoff volume and conduit degree of filling, allowing areas of hydraulic loading and drainage-system stress to be identified. Local climate regulation was assessed using the Urban Multi-scale Environmental Predictor model, specifically SOLWEIG, with inputs including land cover, digital terrain and surface models, building data, canopy information and meteorological variables for summer heat conditions. Outputs included thermal indicators, such as mean radiant temperature and the Universal Thermal Climate Index, which are relevant for assessing outdoor thermal stress and human comfort.
The case shows that, in urban settings, anthropogenic inputs are simultaneously pressures and enabling conditions. Soil sealing, built density and drainage constraints may reduce regulating capacity, whereas tree planting, canopy maintenance, desealing, permeable artificial surfaces, drainage design and spatial planning can guide interventions that enhance realised ES flows. The Urban Greening Plan translates these modelling outputs into priorities for a “fresh, inclusive, resilient and green” city, including shaded streets, an urban river park, desealed areas, mitigation areas and biodiversity-orientated regeneration opportunities (Comune di Trento 2024). Thus, co-production assessments in urban environments require, inter alia, high-resolution spatial data and models capable of representing both ecological processes and built infrastructure. However, separating natural from anthropogenic contributions remains challenging, as thermal comfort and runoff regulation emerge from complex interactions amongst vegetation condition, soil permeability, urban morphology, drainage infrastructure, maintenance and planning decisions.
Triglav National Park, Slovenia: protected landscapes and commons governance
The Slovenian case begins from a different premise: protection status alone does not determine realised ES supply. Triglav National Park (TNP), located in the north-western Alpine Region, is Slovenia’s only national park and covers approximately 84,000 ha. It is divided into three protection zones with different management regimes, ranging from strict nature conservation to areas where sustainable agriculture, forestry, settlements, recreation and tourism are permitted. Forests dominate the land cover, followed by rocky mountain areas, scrublands and permanent grasslands, including meadows and mountain pastures. TNP is, therefore, a multifunctional protected landscape in which conservation objectives must be reconciled with traditional grazing, forestry, tourism, recreation and the management of common lands by agrarian communities (Šmid Hribar et al. 2025a). The case focuses on global climate regulation and biomass (forage) provisioning (Šmid Hribar et al. 2025a), while drawing on a broader protected-area assessment that also considered pollination and nature-based recreation-related ES (Šmid Hribar et al. 2025b).
Co-production in this case is mediated by protection zoning, land tenure and common-land governance. Natural capital includes forests, scrublands, grasslands, high mountain areas, climatic gradients, slope, insolation and vegetation productivity; growing-season length was used as a biophysical indicator of forage-production potential. Human capital is represented by the knowledge and collective practices of agrarian communities’ members, protected-area managers, forestry and conservation experts and researchers. Social capital is central as agrarian communities mediate access, management responsibilities and benefit realisation, while the Park’s zoning framework regulates permissible uses. Manufactured and financial capital are not the primary focus of the assessment, but are implicit in monitoring systems, geospatial datasets, management planning and national or EU support for private and communal land management.
The case combines participatory ES selection and ranking with spatial biophysical indicators. In the broader Slovenian protected-area assessment, ES were selected and ranked through participatory workshops across protected and comparable non-protected areas, while indicator development and ecosystem typology were refined through panel meetings with protected-area representatives and conservation institutions (Šmid Hribar et al. 2025b). Forage supply was assessed as annual grassland biomass increment, expressed in t/ha/year. Potential biomass production was estimated from the length of the growing season and adjusted for rainfall, slope and solar radiation exposure. This approach represents the biophysical potential of grasslands for forage provision, but also indirectly reflects the importance of grazing systems and agrarian-community management for realised use.
Global climate regulation in TNP was assessed through carbon sequestration, using annual Net Primary Production (NPP) as a proxy for carbon uptake, with MODIS NPP data downscaled to 10 m resolution to estimate carbon sequestration across the Park and examine the contribution of common lands managed by agrarian communities (Šmid Hribar et al. 2025a). The authors explicitly note that NPP captures annual carbon uptake rather than long-term carbon storage in biomass and soils and, therefore, represents only one component of the broader climate-regulation service. This limitation is important for co-production assessment because forest and scrubland condition, protection rules and management regimes influence carbon uptake and storage through different mechanisms and temporal scales.
The case demonstrates that nature protection and ES supply are not always equivalent. In the broader area comparison, forage supply did not consistently show higher values in protected areas, unlike pollination and recreation, indicating that outcomes depend on land-use structure and management regimes, as well as protection status (Šmid Hribar et al. 2025b). For carbon sequestration in TNP, forests and scrublands on common lands contribute substantially because of their spatial extent, even where average sequestration capacity is moderate (Šmid Hribar et al. 2025a). Co-production assessment in similar protected areas would, therefore, require indicators that can represent both ecological conditions and management practices. The case also highlights the difficulty of separating ecosystem-derived contributions from human and institutional inputs where commons’ governance, grazing, forest management and conservation rules jointly shape ecosystem condition and realised ES flows.
Discussion
Where explicit co-production assessment is most relevant in integrated ecosystem assessments
The co-production of ES/NCP is well recognised conceptually and across several overlapping framings (Locatelli et al. 2024, Woodhead et al. 2025). Nevertheless, limited operational guidance exists on when and how this recognition should provide information for assessment practice. The expert-elicited ET × ES matrix provides an initial decision-orientated basis for answering the first research question, i.e. explicit assessment of natural and anthropogenic inputs is most relevant where realised ES flows cannot be interpreted reliably from ecosystem condition or ecosystem potential alone and are strongly mediated by management, infrastructure, access, knowledge or institutions. This is directly relevant for integrated ecosystem assessments, where decisions about ecosystem extent, condition, ES supply, use, beneficiaries and values must be connected within a transparent assessment chain (Lange et al. 2026).
Intermediate verdicts dominated the matrix, suggesting that, within the panel’s European reference context, anthropogenic inputs were judged to have at least some relevance for most assessed combinations, but also that detailed disentangling is not always necessary. The strongest pattern is the high relevance of anthropogenic inputs in Urban and Cropland ETs and in selected Woodland and Forest combinations. This is consistent with the capital-based literature, which shows that many ES/NCP are generated through combinations of ecological structures and anthropogenic capital, including labour, knowledge, infrastructure, institutions and finance (Palomo et al. 2016, Jones et al. 2016, Kachler et al. 2023). Urban ES are co-produced in highly hybrid systems in which vegetation, soils and water interact with buildings, drainage networks, transport corridors, planning rules and maintenance regimes. Cropland ES similarly depend on topography, soils, climate, biodiversity and water, but realised flows are strongly shaped by labour, irrigation, fertilisers, pesticides, machinery, agronomic knowledge, subsidies and market access. These findings indicate that, in highly managed ETs, there may be added value in explicitly accounting for the anthropogenic inputs that enable, intensify or redistribute realised ES/NCP flows, particularly where assessment and decision-making processes can benefit from distinguishing ecosystem-derived contributions from human-mediated inputs. Such information can also reveal cases where agricultural ES flows are maintained mainly through input-intensive management, with associated externalities, trade-offs or weak alignment between land use and site-specific ecological conditions. This can support more context-sensitive land-management decisions that better harness local natural potential and reduce dependence on external inputs.
A second, cross-cutting pattern is the generally high relevance of anthropogenic inputs for cultural and appreciation services across different ETs (e.g. recreation-related services, visual amenity, education, scientific and research and spiritual, artistic and symbolic services). As these ES flows are realised only through human presence, perception and meaning, they depend not only on ecosystem potential or capacity, but also on access, interpretation, social practice, identity, institutions and value attribution, making anthropogenic contributions intrinsic to them (Fischer and Eastwood 2016, Raymond et al. 2018). This supports meaning-based and NCP-orientated framings that regard beneficiaries, knowledge systems and cultural interpretation as partly constitutive of realised ES flows (Fischer and Eastwood 2016, Kadykalo et al. 2019, Woodhead et al. 2025). By contrast, lower or more intermediate relevance verdicts for ETs, such as Open Ocean and Sparsely Vegetated Land and for ES related broadly to atmospheric regulation, suggest that explicit co-production assessment should be applied selectively. These findings should be read as structured guidance based on expert judgement from an informed panel, not as statistically representative evidence. In other words, relevance scores and trust tiers indicate where co-production assessment is likely to add value, not universal magnitude or causal strength of anthropogenic contributions.
Which capitals matter and where they enter the ES assessment chain
The second research question of the study is to locate different capitals within the ES assessment chain. The capital- and cascade-based framings locate natural capital as the basis of ecosystem potential, i.e. ecosystem extent, condition, structure, function and ecological processes determine ecosystem potential or capacity. However, realised ES flows frequently depend on anthropogenic capital entering at different chain stages (Palomo et al. 2016, Jones et al. 2016, Lavorel et al. 2020).
The Locatelli et al. (2024) distinction between ecosystem-level and flow-level co-production is central for integrated ecosystem assessments. This distinction helps avoid conflating ecosystem capacity with realised ES flows and clarifies when natural and anthropogenic contributions can be separated, compared or only jointly represented. It also sharpens sustainability interpretation, for example, substituting anthropogenic capital for degraded natural capital may maintain short-term flows, but can increase externalities, reduce resilience and redistribute benefits and burdens inequitably (Palomo et al. 2016, Lavorel et al. 2020).
The case study examples illustrate this chain logic in assessment practice under varying contexts and management regimes. In the German and Cypriot croplands, physical and financial capital (i.e. machinery, irrigation, terracing and purchased inputs) dominate flow-level mobilisation and are readily quantified. In the Slovenian commons and the Greek restoration case, social (and institutional) capital (i.e. land tenure, grazing rules, protected-area governance and Forest Service responsibilities) is decisive, yet enters largely as governance conditions that are acknowledged, but seldom measured (Isaac et al. 2022). This asymmetry, in which manufactured capital is quantified while social and human capital remain under-measured, is a recognised bias in the empirical evidence base (Kachler et al. 2023) with direct consequences for comparability and decision support. Capitals are, therefore, neither equally important nor equally measurable across contexts and integrated assessment design should identify the chain stage and capital types that materially shape the focal ES flow.
Matching assessment approaches to ecosystem types, services and decision contexts
The third research question concerns method choice. The evidence synthesis and the six method families grouped in the literature review section support methodological pluralism rather than a universal co-production method. This follows Woodhead et al. (2025), who show that co-production research is organised through multiple framings: input-focused, actor-focused, context-focused and discipline-focused. For integrated ecosystem assessments, these framings should be treated as complementary analytical entry points.
In practical terms, indicator and typology approaches are useful for screening and comparability (as in the northern German croplands case, Bethwell et al. (2021)); spatial, supply–use and cascade approaches link ecosystem potential, flows and beneficiaries; production-function, econometric (Grammatikopoulou et al. 2020, La Notte 2022) and cost–benefit approaches (as in the Cypriot terrace case) can quantify marginal contributions or economic viability where inputs and outputs are measurable (Lavorel et al. 2020, Locatelli et al. 2024); process-based and spatial decision-support approaches suit restoration, adaptation and hybrid urban systems (as in the Peloponnese and Trento case-studies; Tzamtzis et al. (2023), Geneletti et al. (2025)) and participatory or governance-orientated approaches are essential where access, meanings, institutions and benefit-sharing shape realised flows (as in the Slovenian commons; Šmid Hribar et al. (2025a)).
The unifying principle for selecting a methodological pathway is fitness for purpose. The choice of method for operationalising ES co-production should be proportional to the decision problem, the relevant chain stage, the dominant capitals and the available data and should match the required analytical depth (Balvanera et al. 2022, Barton et al. 2024). As a guiding signal, practitioners may use the relevance matrix as an initial screening tool, along with the method families and cases presented in this study.
Implications for integrated ecosystem assessment, ecosystem accounting and decision-making support
The recently proposed Framework for Integrated Ecosystem Assessment (FIEA) provides a modular structure, organised around six main assessment phases – Frame, Scope, Design, Assess, Share and Act – linking stakeholder engagement, ecosystem condition, ES assessment, ecosystem accounting and the translation of outputs into action (Lange et al. 2026). This study adds a practical co-production layer to FIEA-based assessments, providing information for choices in three phases: (i) whether explicit co-production assessment is relevant for a given ET × ES combination (Frame and Scope, informed by the relevance matrix); (ii) which capitals and chain stages to represent (Design, informed by the conceptual review and case evidence); and (iii) how elaborately to assess them and with which methods or indicators (Design and Assess, informed by the method families). This decision-oriented layer should not be conflated with the co-creation process through which the FIEA itself was developed.
The overarching implication is one of proportionality. The findings point not towards exhaustive measurement and quantification of all natural and anthropogenic inputs, but towards identifying the minimum set of contributions and, where relevant, associated uncertainties as highlighted by Walther et al. (2025) – that must be represented to avoid misleading assessment results. Detailed disentangling of natural and anthropogenic contributions adds most value where anthropogenic inputs materially shape realised flows, where misattribution would distort decisions on accounting, restoration, nature-based solutions or benefit distribution (Jones et al. 2016, Isaac et al. 2022) and where data, time and resources justify the effort, ensuring fit-for-purpose evidence rather than uniform analytical intensity. This also has a distributional implication, i.e. when anthropogenic inputs are ignored, assessments may obscure who maintains, funds, enables or is excluded from realised ES flows. In lower relevance or low-trust ET × ES combinations, a qualitative statement, sensitivity analysis, uncertainty statement or transparent boundary choice may be more appropriate than a full co-production assessment.
Limitations and future research
The study’s findings are bound by certain limitations that need to be acknowledged. First, the relevance matrix is based on structured expert judgement from an informed, but non-representative panel; the final verdicts should, therefore, guide prioritisation rather than be interpreted as statistically representative evidence. Expert coverage also varied amongst ETs, with Shelf representing the clearest case of limited coverage; its retention depends on the adopted lenient knowledge-gap rule, which limits the strength of any ET-level inference for Shelf. Secondly, the five case studies are illustrative rather than confirmatory and were selected to demonstrate methodological pluralism across (European) contexts. Existing empirical evidence remains concentrated on croplands, urban systems and managed forests (Kachler et al. 2023), which limits comparability across ETs and ES and constrains decision support. Thirdly, several capital types remain difficult to measure consistently, particularly social capital in cultural, governance-mediated and access-dependent ES. Future integrated ecosystem assessments would benefit from testing the matrix approach across additional regions, improving cross-context comparability, integrating qualitative and quantitative methods and developing indicators that better capture feedbacks, trade-offs, resilience and equity in co-produced ES/NCP systems.